Acoustic Testing Model - Perceived Spatial Quality of Sound Processing and Reproducing Equipment

Summary

Current models of audio testing measure `basic audio quality', using subjects to judge any difference. This leads to a perceived magnitude of distortions and a subjective acceptability.

The QESTRAL acoustic testing model looks at any and all difference in the spatial attributes.

Benefits

• Reduce the number of listening tests, which are resource intensive.
• Evaluate the effects of different forms of spatial rendering
• Algorithms for optimisation of sound reproduction systems
• Measure the spatial impairment introduced by audio processors, transmission systems or codecs.
• Extend existing standard quality models
• Increase efficiency of industrial development process
• Online check of broadcast quality

Applications

• Audio Standards

Availability

Available for licence

Demonstrator Available

IP status

Patent Pending

Technical

This has been developed over a 3 year period in collaboration between the University of Surrey, BBC R&D and Bang & Olufsen.

Two multi-stimulus listening tests were used to collect listeners¿ judgements of spatial quality to calibrate the regression model. The test used 31 listeners, 43 degradations, 3 types of programme material and 2 listening positions.

The model was calibrated using least-squares regression, with a fixed constraint at the top of the scale. A leave-one-out cross-validation gave R2 = 0.78 and a root-mean-squared error of prediction (RMSEP) of 12.0%.